Effects of norms, warm-glow and time use on household recycling

Discussion Papers No. 389, September 2004 Statistics Norway, Research Department

Bente Halvorsen

Effects of norms, warm-glow and time use on household recycling

Abstract:

The aim of this paper is to quantify the relative importance of motivations based on warm-glow, social and moral norms and cost of time used recycling on household recycling efforts. We also test for crowding-out of intrinsic motivations when recycling is perceived as mandatory. We find that the most important variable increasing household recycling efforts is agreeing that recycling is a pleasant activity in itself, which may be interpreted as a warm-glow effect. The most important variable reducing household recycling is the opportunity cost of time spent recycling. We find no evidence of crowding-out of intrinsic motivation when recycling is perceived as mandatory. On the contrary, we find that governmental legislation increases household recycling efforts on most materials.

Keywords: Contingent valuation method, cost of time, household recycling, moral and social norms, ordered probit, simultaneous estimation of discrete- and continuous choice, warm-glow, crowding out.

JEL classification: C20, C35, D12, D60, H42, H79, Q29

Acknowledgement: The author is grateful to Gorm Kipperberg, Iulie Aslaksen, Karine Nyborg and Erling Holmøy for valuable comments on previous versions of this paper.

Address: Bente Halvorsen, Statistics Norway, Research Department. E-mail: btl@ssb.no

Discussion Papers comprise research papers intended for international journals or books. A preprint of a Discussion Paper may be longer and more elaborate than a standard journal article, as it may include intermediate calculations and background material etc.

Abstracts with downloadable Discussion Papers in PDF are available on the Internet:

http://www.ssb.no

http://ideas.repec.org/s/ssb/dispap.html

For printed Discussion Papers contact:

Statistics Norway

Sales- and subscription service NO-2225 Kongsvinger

Telephone: +47 62 88 55 00

1. Introduction

If household behavior were motivated by economic self-interest only, voluntary contributions to public goods would be negligible. However, in real life, seemingly unselfish behavior is frequently observed and the extent of such non-economically motivated activities may be both time-consuming and costly. One example is Norwegian households' recycling activities. In Norway, household recycling is mainly voluntary with few economic incentives. During the late nineties, several municipalities started recycling programs for their inhabitants. Recycling bins for paper, organic waste etc. was placed at the curb and recycling containers for glass, metal, clothes, etc. were placed at public locations. The inhabitants may use these facilities free of costs with no official sanction possibilities for respondents who do not want to recycle. The programs vary in extension between municipalities, both with respect to the number of materials included and the collecting facilities (curbside, local or central drop-off stations for recycled materials). In 1998, half of the municipalities offered volume- based pricing linked to the remaining waste. Unfortunately, we do not know how much the fees differ, nor do we know how many households that use differentiated fees (Statistics Norway, 2001). Since the system was new in 1999, which is the year analyzed here, we have reason to believe that the number of households making use of the scheme with differentiated fees was low. Despite this, Norwegian households took on a considerable effort sorting, folding, washing, carrying and transporting sorted waste. Thus, the main motivations for household recycling efforts were presumably non-economic.

The share of total waste recycled by households in 1998 was 50 percent for paper and cardboard, 62 percent for glass, 3 percent for plastic and 43 percent for metal (Statistics Norway, 2001, table G10).

So, why do households recycle when it is voluntary, the economic incentives are few and governmental sanctions are limited? Several hypotheses explaining the occurrence of seemingly non- selfish behavior have emerged in the psychological literature (se e.g. Andreoni, 1990, Rabin, 1998, Frey, 1994, Blamey, 1998, Deci and Ryan, 1985, Festinger, 1957, Schwartz, 1970). The literature

often distinguishes between intrinsic and extrinsic motivation. Intrinsically motivated behavior is mainly founded on the individual's view of his/her own worth and self-respect. Examples of intrinsically motivated behavior may be moral norms in the form of a rule of conduct, like the golden rule, or altruistic preferences. The pleasure of giving, also referred to as "warm-glow", is another hypothesis for intrinsically motivated behavior (see e.g. Andreoni, 1990). The argument is that of impure altruistic preferences, as individuals gain utility from contributing to a just cause. The existence of social norms give rise to extrinsically motivated behavior, as individuals may recycle to keep up appearance and gain respect in the community, or to express their attitude towards environmental issues (Blamey, 1998). Extrinsic motivations may also come in the form of economic incentives provided by the government to promote household recycling activities, e.g. through the price mechanism or making recycling mandatory. The problem with introducing economic incentives is that they may crowd-out the intrinsic motivation, as they may reduce self-image (Brekke et al., 2003) and/or raise questions of income distribution and fairness (Frey, 1997). Another explanation for seemingly unselfish behavior is that consumers may have multiple preferences, both concerning private welfare and the welfare of the community (Margolis, 1982, Nyborg, 2000, Sen, 1977). Much of the literature on norms is theoretical, but some empirical studies have been made based on real life data or on experiments (Blamey, 1998, Tögersen 1994, Davis and Holt, 1993, Roth et al., 1991, Eisenberger and Cameron, 1996, Deci, Koestner and Ryan, 1985).

The main focus in the empirical literature on household recycling is how various recycling programs and differentiated tariffs affect household recycling behavior (see e.g. Hong et al., 1993, Jenkins et al., 2000, Jankus et al., 1996, Tiller et al., 1997). Some empirical studies have been made examining the motivation for household recycling efforts and environmental behavior (Vining et al., 1990, and 1992, Pardini and Katzhev, 1984, Hornik et al., 1995, Hopper and Nielsen, 1991, De Young, 1986). The main aim of these studies is to give advice on how to increase the voluntary recycling effort. One empirical study discussing the motivations for household recycling as well as the effects of economic

incentives is Tögersen (1994). Tögersen finds evidence of re-framing and crowding-out of moral norms when economic incentives are introduced in the form of differentiated garbage fees.

While much of the existing empirical work is based on theoretical models of recycling as a household production activity, only a few studies explicitly recognize the time use and its associated cost in the household’s recycling decisions. Two notable exceptions are Jakus et al. (1996) and Hong et al.

(1993). The former analyzes drop-off recycling in a rural community while the latter analyzes curbside recycling in a metropolitan area. Both studies infer the wage rate from income data as a measure of the household’s opportunity cost of time.

The aim of this paper is to quantify the importance of both intrinsic and extrinsic motivations for household recycling based on warm-glow, social and moral norms, in addition to the cost of lost leisure. We use data from a survey conducted in 1999, asking 1162 respondents drawn from the Norwegian population about their recycling efforts, their motivation for recycling and their willingness to pay (WTP) for leaving the recycling to others. In the data, we have information about household recycling activities on six materials: paper, cardboard, plastic, metal, glass and organic waste excess of yard waste. We also have information on various household characteristics such as household income and work hours. Descriptive analysis of household recycling efforts based on this data and a discussion of the motives is given in Bruvoll et al. (2002).

The main contribution of this analysis is how we model the opportunity cost of lost leisure using the properties of the compensating variation (CV) for reducing time spent recycling. Most studies assume the opportunity cost of time used recycling equals the wage rate (se e.g. Jakus et al., 1996, or Hong et al., 1993). Whether this is the case in real life is an empirical question, depending on the household’s preferences and behavior. To estimate the compensating variation (CV) for all costs and benefits of household recycling efforts, we use the stated willingness to pay (WTP) from both an open-ended

(OE-) and a discrete choice (DC-) contingent valuation (CVM) question of leaving recycling to others.

The response to the OE- and DC-CVM questions are estimated simultaneously, modifying a method suggested by Kealy and Turner (1993). This estimation is used to predict the opportunity cost of time used recycling. We also model how social and moral norms and other intrinsic and extrinsic motivations, in addition to the opportunity cost of time used recycling and household characteristics, affect household recycling behavior. This is done by estimating an ordered probit model on each of the six materials.

The paper is organized as follows: In section 2, we model household recycling behavior and discuss how it is affected by altruistic behavior, social and moral norms, and the opportunity cost of time used recycling. Then we discuss how to use the households compensating variation for leaving the household recycling efforts to a renovation firm to predict the opportunity cost of time used recycling.

Then, in section 3, we describe the data and discuss the econometric specification of the model. The results from the estimations are presented in section 4, and in section 5 some concluding remarks are made. Descriptive statistics of all variables used in the estimations are given in the Appendix.

2. The model

Household utility is likely to be influenced by social and moral norms, since how we live up to these norms determine our sense of self-respect and the respect we gain, or sanctions we receive, from the community. We assume that the household gains utility (Uh) from the consumption of a vector of goods (X_h=

{

x_1h,...,x_Mh

}

), leisure (lh), and the consumption of non-economic goods and services such as environmental quality (G), self-respect and respect in the community (Rh) and the warm-glow of contributing to a just cause (WGh), conditional on household characteristics (β_h). The household is also assumed to meet sanctions (Sh), mainly in the form of social sanctions from the community or renovation workers, if the household does not recycle sufficiently. The utility function is given by:

(

_{h h h h h h}

)

h

h U X l R S GWG

U = , , , , , ;β (1)

Household self-esteem (Rh) is assumed to increase, whereas the sanctions (Sh) are assumed to decrease with the quantity of waste recycled by household h (gh). Environmental quality (G) is assumed to increase with the total quantity of recycling in society, defined as the sum of recycled quantity from household h (g_h) and from other households (g_−h). The warm-glow (WGh) from contributing to a just cause is assumed to increase with the quantity of waste recycled by the household. Finally, we assume that the time spent recycling (eh) increases with recycled quantity. We have:

( ) ( )

( )

( ) ( )

^{0 0}

0 0 0

′>

=

′>

=

>

′ +

=

′<

=

′>

=

−

e where g

e e

G W where g

WG WG

G where g

g G G

S where g

S S

R where g

R R

h h h

h h h

h h

h h h

h h h

(2)

2.1 The household’s optimization problem

In this model, we assume that household recycling is a short-term decision whereas the labor-supply location choice is viewed as a more long-term decision. This means that the labor supply is given in the short term, which has several implications. Fist, since the working hours are given, household income (Yh) is given. We assume that the household uses all money income on consumption of goods:

∑

=

=

M

i i ih

h px

Y

1

, where pi is the price on good i. Second, the total numbers of hours that can be used at

leisure or in household production (LTh) will also be given (equal to the number of waken hours excess of working hours). This time is allocated between recycling (eh) and other household production or leisure activities, hereafter referred to as excess leisure time (l_h): LT_h =l_h+e_h. So, conditional on the labor supply, we analyze how leisure is divided between time spent on recycling and excess leisure.

The household is assumed to maximize utility with respect to the consumption of all goods (Xh), excess leisure time (lh) and recycled quantity (gh), subject to the time and money budget. This maximization problem gives rise to the following Lagrange function:

( ) ( ) ( ) ( )

( )

h h

{

h

( )

h h h

}

M

i i ih h h h h

h h h h h

h U X l Rg S g G g WG g px Y e g l LT

L − + −

⎪⎭

⎪⎬

⎫

⎪⎩

⎪⎨

⎧ −

−

=

∑

=

ξ η

β

1

, , , , , ; (3)

where η_h is the Lagrange multiplier for the money budget and ξ_h is the Lagrange multiplier for the time budget. This optimization problem gives the following first order conditions:

, 0 ...,

, 1 , 0 0

′≡

′−

∂ + ∂

∂ ′ +∂

∂ ′ +∂

∂ ′

=∂

∂

∂

=

∀

≡

∂ −

=∂

∂

∂

≡

∂ −

=∂

∂

∂

h h h

h h

h h

h h

h h h

i h ih

h ih

h

h h

h h

h

e G WG W G U G S U S R U R U g L

M i x p

U x L

l U l L

ξ η

ξ

(4)

When considering excess leisure time, the household evaluates the utility effects of increased lh for a given time budget. When deciding on consumption of good i, the household evaluate the direct utility effects against the cost of increased consumption for a given money budget. Finally, when deciding on recycled quantity, the household evaluates the utility effects of increased respect, feeling of warm- glow, increase in environmental quality and reduced sanctions for a given time budget. Thus, the Lagrange multiplier for the time budget (ξ_h) equals the utility effects of increased recycling per time unit and may thus be interpreted as the opportunity cost of time used on recycling. From these first order conditions we see that in optimum, the utility of a marginal increased excess leisure time must equal the utility effects of a marginal increase in time used recycling.

Solving all first order conditions of this maximization problem except one, using that both the time and money budget must be fulfilled in optimum, gives household recycling activities, excess leisure time and consumption of goods as a function of all prices (P), household income (Yh), total leisure time (LTh), recycling by other households (g_−h) and the opportunity cost of time used recycling (ξ_h), conditional on household characteristics (β_h).

( )

( )

(

h h h h h

)

ih ih

h h h h h h h

h h h

h h h h

g LT Y P x x

g LT Y P L l

g LT Y P g g

β ξ

β ξ

β ξ

; , , , ,

; , , , ,

; , , , ,

−

−

−

=

=

=

(5)

The opportunity cost of time spent recycling in optimum (ξ_h^*) is found by solving all first order conditions with respect to the Lagrange multiplier for the time budget. This gives the opportunity cost of time as a function of all prices, household income, total time spent at leisure, and recycling by other households, conditional on household characteristics:

(

_{h h h h}

)

h

h ξ PY LT g β

ξ^*= , , , ₋ ; (6)

Inserting the opportunity cost of time in optimum into equation (5) gives the optimal recycling of waste, excess leisure time and consumption of goods for the household: g_h^*=g_h

(

P,Y_h,LT_h, g_−h;β_h

)

,

(

_{h h h h}

)

h

h L PY LT g

l^*= , , , ₋ ;β , and x_ih^* =x_ih

(

P,Y_h,LT_h, g_−h;β_h

)

. Inserting this into the utility function in (1), gives the household’s indirect utility as a function of all prices, household income and optimal recycling, conditional on household characteristics:

( ) ( ) ( ) ( )

(

h h h h h h h h h h h

)

h

h V P Y LT g R g S g G g WG g

V = , , , ₋ , ^* , ^* , ^* , ^* ;β (7)

2.2 The cost of time used recycling

The optimal recycling activity depends on all factors determining utility from recycling, such as respect, sanction if the household do not comply, environmental quality, warm glow of contributing and the opportunity cost of time spent recycling in optimum. In this paper, we want to illustrate how these factors affect household recycling behavior. In particular, we are interested in how the opportunity cost of time used on recycling affects recycling behavior. In order to find a measure for the individual household's opportunity costs of time, we use the household compensated variation for leaving recycling to others (CVh).¹

The household's CV is defined as the reduction in expenditures necessary to sustain utility after the recycling effort is left to others. Thus, the CV is the reduction in income that makes the household indifferent between doing the recycling themselves and leaving recycling to others. If a renovation firm takes over the sorting of household waste, the households gain utility since they no longer need to spend time recycling (eh(0) = 0) and it may use all its available leisure time as excess leisure time (lh = LTh). Additionally, the household does no longer feel a pressure to recycle in order to avoid sanctions, nor do they feel the sense of warm glow and/or self-respect from contributing to a just cause when the recycling is left to others (Sh(0) = Rh(0) = WGh(0) = 0). The CV for leaving household recycling to others may either be positive or negative depending on household preferences concerning these factors. For simplicity, we assume that the prices (P) are not affected by the household’s recycling decision, and they are thus left out of the indirect utility function. Inserting the time budget in the indirect utility function and using that lh = LTh when the household does not recycle (eh(0) = 0), the CV is defined by:

( ) ( ) ( ) ( ) ( )

(

, ^* + ^*, ^* , ^*, ^* , ^*;

)

−

(

− , ,0,0, ,0;

)

=0

=

∆V_h V_h Y_h e_h g_h l_h R_h g_h S_h g_h Gg_h WG_h g_h β_h V_h Y_h CV_h LT_h G β_h (8)

where G is the environmental quality generated by the total amount of waste recycled by the firm. For simplicity, we assume that the same amount of waste is sorted regardless of who is doing the recycling, the household or a renovation firm. Thus, the environmental quality is equal in the two situations: G=G

(

gh^*+g₋h

)

. Solving this with respect to the CV gives the household compensating variation for leaving recycling to others as a function of household recycling in optimum, household income and total leisure time available, conditional on household characteristics:

(

h h h h

)

h

h CV g Y LT

CV = ^*, , ;β (9)

We decompose the welfare effect of leaving recycling to others (∆V_h) into the welfare effects of warm glow, reduced sanctions and self-respect (∆V_h^S,R,WG) and the welfare effects of reduced time spent recycling (∆V_h^e), that is ∆V_h =∆V_h^S,R,WG+∆V_h^e, defined by:

( ) ( )

(

h h h h

)

h

(

h h h

)

h

h h

h h h h h h h h h h h h

G LT Y V G

l e Y V

G l e Y V WG

G S R l e Y V V

β β

β β

; 0 , , 0 , 0 , ,

; 0 , , 0 , 0 , ,

; 0 , , 0 , 0 , ,

; , , , , ,

*

*

*

*

*

*

*

*

*

− +

+

+

− +

=

∆ (10)

where the first utility difference equals ∆V_h^S,R,WG and the second difference equals ∆V_h^e. Using the intermediate-value theorem, a welfare change equals the marginal utility of income, evaluated at some intermediate income Yh∈

(

Y_h,Y_h−CV_h

)

, times the compensating variation: ∆V_h=CV_hV_hY′ (see e.g.

Johansson, 1993, equation 3.5). Using this and equation (10) we are able to decompose the total CV into the CV for reduced time recycling (CV_h^e) and the CV for all other factors (CV_h^S,R,WG), defined by:

1 For a discussion of the Compensating Variation, see Varian (1997), Mas-Colell et al (1995) or Mitchell and Carson (1989).

e h WG R S h hY

e h WG R S h

h CV CV

V V

CV V = +

′

∆ +

=∆ ^{, , , ,} (11)

We see from equation (11) that the CV of time spent recycling equals the total CV minus the CV for the utility effects of warm-glow, respect and sanctions. Thus, if we assume that the household does not receive utility from respect, warm glow or sanction, the total CV equals the CV for time spent recycling since the total amount of waste recycled, and thus the environmental quality, is equal in the two situations. If leaving recycling to others affects utility due to warm-glow or respect from contributing, the total CV differ from the CV of time used recycling. In our estimations, we will use the property that the CV of time used recycling measures the opportunity cost of time used recycling (ξ_h) given the households recycling effort (see equations 4 and 11). That is, we use the household's stated maximum willingness to pay to leave their recycling to others as a proxy for the opportunity cost of time used recycling.

3. Econometric specification

3.1 The data

The data used in this analysis is part of a survey conducted by Statistics Norway in November and December 1999. Our sample includes 1162 respondents between the age of 16 and 79 years old drawn from the Norwegian population. The response rate was slightly less than 60 percent of the gross sample of 2000 respondents. For 76 percent of the sample, the respondent was interviewed in person in their own home, while the remainders of the interviews were conducted by telephone.

In the survey, the respondents were asked questions whether they recycle all, most of, some or none of six different materials: Paper, cardboard, glass, metal, plastic and organic waste excess of yard waste.

respondents were told that a waste disposal firm would take over the recycling (“Suppose that a recycling company can utilize your garbage. A new technology allows central source separation such that the environmental impact is the same. The company will collect your garbage at your home.”), and asked if they would be willing to make use of the offer by the waste disposal firm if it came free of charge (“Would you use this service if it did not cost you anything?" Or do you prefer to do your own recycling?”). Then, the respondents who would make use of the firm were asked two Contingent Valuation questions.² First, they were asked a Dichotomous Choice Contingent Valuation (DC-CVM) question ("Are you willing to pay x NOK yearly to a waste disposal firm so they can take over the recycling effort of your household?" x = 50, 150, 250, 450, 600 NOK).³ Then, they were asked an Open-Ended Contingent Valuation (OE-CVM) question: ("What is the maximum yearly amount you are willing to pay in order for the waste disposal firm to take over your household's recycling?")

In order to determine the motivation for household recycling efforts, we included questions concerning attitudes towards the recycling activity. The respondents were asked if they agreed, partly agreed, partly disagreed or disagreed to the following statements: I recycle because …. a) "I want to consider myself as a responsible person", b) "I want others to consider me as a responsible person", c) "I perceive the recycling effort as imposed by the government", d) "Recycling is a pleasant activity in itself", e) "I want to do what I whish others to do" and f) "I want to contribute to a better environment".

We also asked the respondents if they thought recycling contributed to a better environment and if they would think of it as an advantage or a disadvantage if the municipality increased the extent of the voluntary recycling programs. The survey also includes routine questions concerning household in- come and labor supply, and how densely populated the respondents place of residence is. Finally, we

2 See e.g. Mitchell and Carson 1989 for more information on the Contingent Valuation method.

3 One NOK is approximately $0.12 US.

have information concerning the number of household members and the place of residence. (Descrip- tive statistics of all variables included in the analysis is given in appendix table A1.)

In our estimation, we will use the answers to the motivational questions as indications of how norms and warm-glow elements influence household recycling activities. Question a) is interpreted as an indication of intrinsic motivation, whereas b) is an indication of extrinsic motivation to increase the respect gained from recycling. If perceiving recycling as mandatory (question c) reduces household recycling activities, it may be an indication of crowding-out effects of intrinsic motivations. Wanting to contribute to the environment (question f) is interpreted a "warm-glow" effect or the desire to express environmental concerns through recycling. There may, however, be other reasons for wanting to contribute to the environment, e.g. a sense of moral or social obligation to do so. If the motivation is warm-glow, question (f) should indicate a benefit of household recycling. However, if it is motivated by a sense of commitment, it will represent an obligation to recycle. Finding recycling a pleasant activity (d) may be an indication of warm-glow if the reason the activity is perceived as pleasant is the feeling of contributing to a just cause. Wanting to do what I which others to do (question e) is interpreted as an indication of respondents living and acting according to the Golden Rule, which is an indication of moral commitment. Living according to moral norms should represent a cost of recycling if the aim is to avoid sanctions. On the other hand, the household may also gain utility complying with moral norms if it increases self-respect or the respect the household gains in the community. Agreeing that it would be a disadvantage if the municipality increased the extent of the voluntary recycling programs is an indication of social pressure to avoid sanctions. Finally, respondents who recycle despite a belief that it has no effect on environmental quality must have other motivations for their recycling activity, based on warm-glow, social and/or moral norms.

In the survey, only respondents who reported to recycle at least one of the six materials were asked about their recycling efforts. The 7 percent of the sample who did not recycle at all are reported with

missing values for all questions concerning household recycling efforts and moral motivations. Thus, we are only able to explain the behavior of the 93 percent of the sample that engage in some kind of recycling activities. Furthermore, we do not have information about the type of waste collecting services offered by the municipalities, nor do we have information whether the municipalities offer volume-based tariffs or not.

3.2 Household recycling effort

We approximate the household recycling activities (g_h) described in equation (5), by a linear function of household income (Yh), number of household members (Nh), cost of time used recycling (TCh), variables describing the household's attitude towards recycling (Dhk), plus a random error term (ν_h) which is assumed to be normally distributed with a zero mean and a constant variance.

h h h

k

kh k h

Y h TC h N

h N TC Y D X

g =α +α +α +α +

∑

α +ν =α′ +ν

=1 0

* (12)

The attitude variables are used to approximate factors concerning respect, warm-glow, sanctions and environmental concern. The variables included are the respondents' response to questions a) to f) (see the description of the data page 12), if they thought recycling contributed to a better environment and if they would think of it as an advantage if the municipality increased the extent of the voluntary recycling programs so they can recycle more.

Unfortunately, we are not able to observe the household recycling of waste in volume units (g*), as we have only four observable outcomes: The household recycles all, most of, some or none of their recyclable waste. Thus, the observable recycling effort (g~ ) is given by: _h

* 2

2

* 1

1

*

*

~ 3

~ 2

0

~ 1

0

~ 0

h h

h h

h h

h h

g if

g

g if

g

g if

g

g if

g

<

=

≤

<

=

≤

<

=

=

=

κ

κ κ

κ (13)

where κ₁ and κ₂ are the limits where the respondents shift from recycling some to most of the waste, and from most to all of the waste respectively. The household recycling effort may thus be viewed as a latent multinomial decision model (see e.g. in Greene (1993) pp. 672 - 676).

The probability of observing different levels of household recycling efforts are given by:

( ) ( )

( ) ( ) ( )

( ) ( ) ( )

(

g

) (

X

)

P

X X

g P

X X

g P

X g

P

h h h h

α κ

α κ α

κ

α α

κ α

− ′ Φ

−

=

=

− ′ Φ

′ −

− Φ

=

=

− ′ Φ

′ −

− Φ

=

=

− ′ Φ

=

=

2

1 2

1

1

~ 3

~ 2

~ 1

~ 0

(14)

where α′X follows from equation (12). Since we assume that the random error term (ν_h) is normally distributed, the probability of observing g~ is given by the standard normal probability (_h Φ), estimating all parameters in this model (α,κ ) applying an ordered Probit model. The α parameters give the effects on the optimal recycling effort of marginal changes in the explanatory variables.

3.3 The cost of time

The opportunity cost of time used recycling (TCh) in equation (12) is endogenous to the respondents, and must thus be instrumented in order to avoid biased estimates. This is because the covariance between this variable and the error term differ from zero. We estimate an instrument for the opportunity cost of time in optimum (described by equation 6) based the household CV for leaving

both the open-ended (OE-) and discrete choice (DC-) contingent valuation (CVM) questions as a proxy for household CV for the service. Then, we calculate the opportunity cost of time spent recycling as the predicted CV excess of variables indicating benefits and moral and social costs and benefits of recycling.

The household CV for leaving recycling to others (described by equation 9) is approximated by a linear function of household work hours (LSh), household gross income (Yh), the age of the head of the household (Ah), a set of variables indicating moral and social costs (MC_h^k) and benefits (B_h^k) from household recycling and a stochastic error term (ϖ_h) which is assumed to be normally distributed with a zero mean and a constant variance:

h k h k

k B k

h j

j MC h

A h L h Y

h Y LS A MC B

CV =γ +γ +γ +γ +

∑

γ +

∑

γ +ϖ

=

=

5

1 3

1

0 (15)

The variables indicating moral and social costs (MC_h^k) are the respondents response to: "I want to consider myself as a responsible person", "I want others to consider me as a responsible person", "I perceive the recycling effort as imposed by the government", "I want to do what I whish others to do"

and if they would think of it as a disadvantage if the municipality increased the extent of the voluntary recycling programs. The variables indicating the benefits (B_h^k) are the respondents' response to:

"Recycling is a pleasant activity in itself" and "I want to contribute to a better environment". The CV is assumed to be a function of household gross income in order to correct for the household's ability to meet payments. We also include the age of the head of the household, as younger families often have small children, increasing expenditures and the cost of additional time spent on recycling.

Since the respondents were asked both an open-ended and a close-ended WTP question in the questionnaire, we estimate the WTP from both questions simultaneously, applying a method suggested by Kealy and Turner (1993).⁴ In Kealy and Turner, the DC-CVM question was conditional on the OE- CVM question. In our data, the conditioning is the other way around, as the OE-CVM question was asked after the DC-CVM question. Thus, we need to modify the method suggested by Kealy and Turner (1993), making the OE-CVM question conditional on the DC-CVM question.

We assume that the compensated variation from the DC-CVM question (CV^DC) may be written as the sum of the expected CV (µ_DC) and a stochastic error term (ω_DC), where the expected CV is a linear function of a vector of explanatory variables (X): CV^DC=µ_DC+ω_DC=λ′X+ω_DC. Likewise, the CV from the OE-CVM question (CV^OE) is written as the sum of the expected CV (µ_OE) and a stochastic error term (ω_OE):CV^OE=µ_OE+ω_OE=γ′X+ω_OE. These CV functions are given by equation (12).

Furthermore, we assume that the CV^DC and the CV^OE are simultaneously and normally distributed:

(

^,

) (

^{~ µ}DC^,µOE^,σDC^{2 ,σ}OE^{2 ,ρ}

)

OE

DC CV N

CV , where ρ is the correlation coefficient between CV^OE and CV^DC, and σ_OE² and σ_DC² are the variances of CV^OE and CV^DC. In the DC-CVM question, the respondents were asked whether they were willing to pay the cost (C) for leaving the recycling to others or not. We define a dummy variable y that equals one if the CV^DC exceeds the cost C (y = 1 if CV^DC > C), and zero otherwise, that is, if the respondent does not accept the offer. Furthermore, we assume that the respondent's CV in the OE-CVM question is conditional on the answer they gave to the proceeding DC-CVM question. Thus, the conditional distribution of CV^OE is given by:⁵

( ) ( )

⎥⎥

⎦

⎤

⎢⎢

⎣

⎡ ⎟⎟ −

⎠

⎜⎜ ⎞

⎝ + ⎛ −

= , ² 1 ²

| σ ρ

σ σ µ ρ

µ _OE

DC DC OE

OE DC

OE C

N CV CV

f .

4 See also Halvorsen and Sælensminde (1998) for a discussion of the method.

The simultaneous probability of observing the combined answers to the DC- and OE-CVM questions are given by the product of the conditional marginal probability of stating a maximum WTP equal to x and the probability of stating i = yes or no in the DC-CVM question: P(CV^OE = x, y = i) = P(y=i)*P(CV^OE = x | y = i). Applying the simultaneous probability of observing the combined answers to the CVM questions, gives the following Log-likelihood function for the simultaneous choice:

( )

( ) ( )

∑

∑

=

=

⎥⎥

⎥⎥

⎥

⎦

⎤

⎢⎢

⎢⎢

⎢

⎣

⎡

⎪⎪

⎭

⎪⎪

⎬

⎫

⎪⎪

⎩

⎪⎪

⎨

⎧

−

⎟⎟⎠

⎜⎜ ⎞

⎝

⎛ − ′

′ −

− + −

⎥⎥

⎦

⎤

⎢⎢

⎣

⎡

⎪⎭

⎪⎬

⎫

⎪⎩

⎪⎨

⎧

⎟⎟⎠

⎜⎜ ⎞

⎝

⎛ − ′ Φ

−

⎟⎟+

⎠

⎜⎜ ⎞

⎝

⎛ − ′ Φ

−

=

N

h OE

DC OE OE

OE N

h DC DC

X X C

CV

X y C

X y C

L Log

1 2 2 2 2

1

1 1

log 1

1 1

log )

(

ρ σ

σ σ λ ρ γ ρ ϕ

σ

σ λ σ

λ

(16)

This Log-likelihood function is maximized applying the MINIMIZE-procedure in Limdep, estimating all parameters simultaneously applying a FIML-procedure.⁶ This estimation yield two estimates of the WTP, one from the OE- and one from the DC-CVM equation. We use the mean WTP from the two WTP-questions for each individual household as a proxy for household's opportunity cost of time in the ordered probit estimation of equation (12).

Here, we assume the respondents statements to the two CVM questions reflects their compensated variation of leaving recycling to others. In the literature on the contingent valuation method, one of the main controversies is whether different elicitation formats under- and/or overstates the respondents true CV. In the theoretical literature, it is often assumed that the open-ended format gives the respondents incentives to overstate their true WTP (see e.g. Arrow et al., 1993, for a discussion).

Comparative empirical analyses, on the other hand, mainly obtain higher WTP estimates using the DC-CVM format as compared to the OE-CVM format (see e.g. Halvorsen and Sælensminde, 1998). In

this analysis, we apply the responses to both questions in order to extract as much information about the households WTP as possible, and hopefully avoid systematical bias in the results as a consequence of the chosen elicitation format.

4. Empirical results

First in this section, we present the estimates from the simultaneous estimation on the WTP statements from the OE- and DC-CVM questions. Then, we present the results from the ordered Probit model on the share of waste recycled on six materials.

4.1. Households' time cost

In table 1, we present the results from the simultaneous estimations on the OE- and DC-CVM questions. In the estimation, household income, work hours, age of the respondent and variables determining the households' motivation and attitudes towards recycling were used as independent variables explaining the stated WTPs from both the OE- and the DC-CVM question. The attitude variables may take four values; disagree = 0, partly disagree = 1, partly agree = 2 and agree = 3 to the statements. In the first column of the table we present the estimated coefficients. In the next column, we present the T-value and in the last column we present the P-value. In the first section of the table, we present the results from the OE-CVM equation, and in the second section we present the estimates from the DC-CVM equation. Finally, in the last section, we present the estimated standard deviations and correlation coefficients.

Table 1: Results from the simultaneous Maximum Likelihood estimations on the willingness to pay for handing over household recycling efforts to others from a dichotomous choice- (DC-CVM) and an open-ended (OE-CVM) contingent valuation question

Variable Coefficient T-value P-value

OE-CVM question (NOK):

Constant 350.38 4.65 0.000

Household gross income (1000 NOK) 0.24 4.48 0.000

Work hours (hours per week) 1.50 2.34 0.019

Age of respondent (years) -7.26 -6.94 0.000

I want to contribute to the environment (0, 1, 2, 3) 21.14 0.85 0.396 I find recycling is a pleasant activity in itself (0, 1, 2, 3) -20.14 -1.83 0.068 I want to consider myself as a responsible person (0, 1, 2, 3) -0.05 0.00 0.997 I want others consider me as a responsible person (0, 1, 2, 3) 6.17 0.50 0.616 I perceive recycling as mandatory (0, 1, 2, 3) -15.42 -1.57 0.117 I which to do what I want others to do (0, 1, 2, 3) 14.40 0.92 0.356 I would think of it as a disadvantage if the municipality increased the

extent of the voluntary recycling programs (0,1) 10.47 0.37 0.709 DC-CVM question (0, 1):

Constant 491.35 3.01 0.003

Household gross income (1000 NOK) 0.30 1.90 0.057

Work hours (hours per week) 3.71 1.75 0.081

Age of respondent (years) -11.03 -3.30 0.001

I want to contribute to the environment (0, 1, 2, 3) 18.75 0.58 0.562 I find recycling is a pleasant activity in itself (0, 1, 2, 3) -17.62 -0.61 0.542 I want to consider myself as a responsible person (0, 1, 2, 3) -0.17 -0.19 0.853 I want others consider me as a responsible person (0, 1, 2, 3) 0.49 0.53 0.593 I perceive recycling as mandatory (0, 1, 2, 3) -19.93 -0.78 0.434 I which to do what I want others to do (0, 1, 2, 3) 18.57 0.63 0.529 I would think of it as a disadvantage if the municipality increased the

extent of the voluntary recycling programs (0,1) 24.70 0.30 0.761 Standard deviations:

Standard deviation from the OE-CVM question 327.06 19.56 0.000

Standard deviation from the DC-CVM question 557.75 4.93 0.000

Coefficient of correlation between the OE- Snd DC-CVM question 0.51 6.83 0.000

Looking at the results, we see that household income increases the WTP for leaving recycling to others significantly, both in the OE-CVM and the DC-CVM question, as the household's capacity to pay increases with income. Work hours have a significant positive effect on both the OE- and DC- CVM question, that is, households with long work hours have a higher WTP for leaving their recycling to others than households with more leisure time. This indicates that the cost of recycling increases with the work hours. This was as expected since leisure time becomes more valuable when the number of working hours increases. We also see that the age of the respondent have a significantly negative effect on the WTP-statements. There might be several reasons for this. One explanation is that younger families often have small children, increasing the marginal cost of time. Another explanation is that older people are more concerned with the moral obligation of contributing to the community, and that younger people are more willing to accept purchasing recycling services.

Looking at the attitude variables, the only significant variable is finding recycling a pleasant activity, which reduces the WTP from the OE-CVM question. For the rest of the attitude variables, none of the coefficients differ significantly from zero. From the last section of the table, we see that the answers to the two WTP questions are not independent, as the estimated correlation coefficient (ρ) differs significantly from zero with an estimate of 0.51. Thus, assuming independence by calculating the WTP for each question separately will not include all information concerning the distribution of the WTP in the data.

In order to obtain an estimate of the households' cost of time used recycling, we use this estimation to predict the household's mean WTP from the OE- and DC-CVM questions net of all attitude variables.

This estimate of the households' opportunity cost of time is then applied in the estimation of the recycling effort on the different materials.

4.2. Household recycling effort

In table 2, we present results from the ordered probit estimations on the share of the six materials recycled by the households. In the first column of table 2, we report the estimated coefficients in the ordered Probit model on the share of paper recycled by the households. In the next five columns, we present the coefficients from the estimated share of cardboard, plastic, metal, glass and organic waste.

The dependent variables may take four values: Recycle none = 0, some = 1, most = 2 and all = 3 of the household's waste. In all estimations, the following variables were used as explanatory variables:

household income, the estimated time costs, the number of household members, age of the respondent, living in a major city, living in a block of flats, variables determining the households' motivation and attitudes towards recycling, if they thought recycling contributed to a better environment and if they think it is an advantage for them if the municipality increased the extent of the voluntary recycling programs so they can recycle more. The attitude variables may take four values: disagree = 0, partly disagree = 1, partly agree = 2 and agree = 3 to the six statements (se statements a) -f) in the discussion of the data in section 3.1). In the last part of the table, estimates for κ₁ and κ₂ are included.

Table 2: Results from an ordered Probit estimation on the share of paper, cardboard, plastic, metal, glass and organic waste recycled by households. ^a

Paper

Card-

board Plastic Metal Glass

Organic waste

Constant 1.3526 ** 0.1321 -1.3767 ** -0.2702 -0.1333 0.2338

Household gross income (1000 NOK) 0.0007 ** 0.0002 -0.0001 0.0003 0.0005 ** 0.0003 Estimated opportunity cost of time spent recycling (NOK) -0.0015 ** -0.0010 ** 0.0000 -0.0008 ** -0.0011 ** -0.0012 **

Number of household members (N) -0.0060 0.0557 * 0.0024 0.0150 0.0175 -0.0029 Age of the respondent (years) -0.0007 -0.0097 ** -0.0017 0.0002 0.0044 -0.0099 **

Major city (0,1) 0.0047 -0.1106 -0.1361 -0.3958 ** -0.2499 ** -1.0889 **

Living in a block of flats (0,1) 0.1796 -0.1464 -0.2473 * -0.1495 0.1461 -0.3535 **

I want to consider myself as a responsible person (0, 1, 2, 3) 0.0188 0.0425 0.0297 0.1405 ** 0.0548 -0.0094 I want others consider me as a responsible person (0, 1, 2, 3) 0.0138 0.0138 0.1038 ** 0.0390 0.0111 0.0684 * I perceive recycling as mandatory (0, 1, 2, 3) 0.0117 0.0015 0.1385 ** -0.0237 -0.0440 0.1734 **

I find recycling is a pleasant activity in it-self (0, 1, 2, 3) -0.0217 0.0958 ** 0.0981 ** 0.0960 ** 0.0452 0.0671 * I like to do what I want others to do (0, 1, 2, 3) 0.1256 ** 0.1579 ** 0.1022 0.0868 * 0.0706 0.0103 I want to contribute to the environment (0, 1, 2, 3) -0.0918 0.0191 0.1097 -0.0363 0.1524 * 0.0482 I believe recycling contributes to a better environment (0,1) 0.0194 0.2935 ** 0.1601 0.2050 * 0.4221 ** 0.1046 I think it is an advantage if the municipality increases the extent of

voluntary recycling programs so I can recycle more (0,1) 0.2089 ** -0.0612 -0.0934 -0.2092 ** -0.0525 -0.1833 **

κ1 0.2428 ** 0.3129 ** 0.2170 ** 0.3102 ** 0.2966 ** 0.1226 **

κ2 0.8898 ** 0.7071 ** 0.5054 ** 0.5793 ** 0.7598 ** 0.2922 **

a) ** implies that the coefficient differ significantly from zero with a probability of falsely rejecting the zero hypothesis at less than 5 percent. * implies that the coefficient differ significantly from zero with a probability of falsely rejecting the zero hypothesis at less than 10 percent.

Looking at the attitude variables, their importance on household recycling efforts varies between materials. For instance, respondents who are more concerned with self-respect, recycle more metal than other respondents. Respondent who are concerned with the respect they receives from others recycle more plastic and organic waste than others. Even if recycling is essentially voluntary in Norway, some respondents perceive it as mandatory. If the respondent perceives recycling as mandatory, it increases household recycling of plastic and organic waste significantly while the coefficient is insignificant for the other materials. The estimated coefficient is negative on the recycling of paper and glass, but these coefficients are not significant. Thus, we do not find significant evidence of crowding-out of intrinsically motivated behavior in our data. Respondents who find recycling a pleasant activity recycle significantly more of cardboard, plastic, metal and organic waste, and respondents agreeing to the Kantian rule recycle more of paper, cardboard and metal than other respondents. Wanting to contribute to a better environment only had a significantly positive effect on the recycling of metal. Respondents who believe that recycling contributes to a better environment recycle significantly more of cardboard, metal and glass than other respondents. Finally, respondents who think it is an advantage if the municipality increased the extent of the voluntary recycling programs so they can recycle more, recycle significantly more of paper and less of metal and organic waste than other respondents. The reason for the negative signs of the coefficient may be that these are respondents with a high motivation for recycling who lives in communities that do not offer recycling programs for marginal materials such as metal and organic waste.

If we take an overall look at table 2, we see that the attitude variables have the largest impact on the recycling of the more marginal materials such as plastic and organic waste, and less on the recycling of paper and glass, where most municipalities has well-functioning recycling programs. We see that especially for the recycling of paper, which is the most common material for curbside recycling programs, only agreeing to the Kantian rule increases the effort significantly. For cardboard, enjoying the recycling activity in itself and the belief that recycling increases environmental quality also increase the effort, probably since the recycling of cardboard involve a lot of cleaning and folding of

milk cartons. We also see that enjoying the recycling activity in itself have a positive effect on the recycling effort of most waste fractions with the exception of paper and glass, where the effort of recycling is smallest.

5. Concluding remarks

Based on the findings from this study, we see that the most important variables explaining household recycling activities were finding recycling a pleasant activity, which may be interpreted as warm- glow, and the estimated opportunity cost of time spent recycling. Beyond this, the importance of the motivations varies with the nature of recycling the various materials. Furthermore, we find that perceiving recycling as mandatory does not seem to crowd-out other motivations for recycling. On the contrary, it seems like perceiving recycling as mandatory reduces the WTP for leaving recycling to others, increases the recycling effort and strengthens the moral motivation. This indicates that the households gain utility when the local government makes recycling mandatory, which may seem odd.

One explanation may be that governmental legislation may influence the publics' attitude towards recycling, increasing awareness and emphasizing the importance of recycling. Another explanation is that households may expect the total recycling in the community to increase, resulting in a higher environmental quality, when recycling is a joint effort secured by legislation where everybody participates. A third explanation may be a sense of moral obligation to sort waste when recycling is perceived as mandatory. Thus, the respondents may see the alternative of leaving recycling to others as slinking away from their obligations as a citizen, reducing their WTP.

Market-based instruments to stimulate economic incentives are gaining popularity in the environmental policy. For instance, during the late nineties, most municipalities introduced various schemes to stimulate household recycling efforts. These mainly consisted of differentiated garbage fees, curbside recycling programs, and/or local and central drop-off facilities for various materials.

The effectiveness of these incentives may depend on how they affect the motivation for household recycling behavior. In this study, we have not included variables describing the recycling programs

and differentiated garbage fees, as they were not available in our data. We have only tested if perceiving recycling as mandatory crowd out morally motivations for recycling, which we find no evidence of. This does not, however, exclude the possibility that for instance introducing differentiated garbage fees may crowd-out intrinsic motivation, which Thögersen (1994) find evidence of in Denmark. How different recycling programs stimulate the recycling of various materials is the main focus of our future research on this topic.

Appendix

Table A1: Descriptive statistics.

Mean Std.Dev. Minimum Maximum NumCases

Willing to pay the cost (0, 1) 0.36 0.48 0 1 1162

Maximum WTP for leaving recycling to others (NOK) 175.97 276.32 0 2000 1059

Recycling effort on cardboard (0, 1, 2, 3) 1.69 1.31 0 3 1162

Recycling effort on paper (0, 1, 2, 3) 2.41 1.02 0 3 1162

Recycling effort on glass (0, 1, 2, 3) 2.14 1.19 0 3 1162

Recycling effort on metal (0, 1, 2, 3) 1.33 1.35 0 3 1162

Recycling effort on plastic (0, 1, 2, 3) 0.84 1.23 0 3 1162

Recycling effort on organic waste (0, 1, 2, 3) 1.35 1.43 0 3 1162

Household gross income (1000 NOK) 392.96 254.75 0 3000 1036

Work hours (hours per week) 25.83 19.73 0 100 1162

Number of household members (N) 2.79 1.42 1 13 1162

Age of the respondent (Years) 42.46 16.41 15 79 1162

Major city (0,1) 0.25 0.43 0 1 1162

Living in a block of flats (0,1) 0.13 0.33 0 1 1162

I want to consider myself a responsible person (0, 1, 2, 3) 1.98 1.11 0 3 1088 I want others consider me a responsible person (0, 1, 2, 3) 1.17 1.23 0 3 1086

I perceive recycling as mandatory (0, 1, 2, 3) 1.75 1.21 0 3 1093

I find recycling is a pleasant activity in it-self (0, 1, 2, 3) 1.11 1.14 0 3 1094 I like to do what I want others to do (0, 1, 2, 3) 2.48 0.85 0 3 1092 I want to contribute to the environment (0, 1, 2, 3) 2.82 0.50 0 3 1094 I believe recycling contributes to a better environment (0,1) 0.85 0.36 0 1 1162 It is an advantage if the municipality increases the extent of

voluntary recycling programs so I can recycle more (0,1) 0.34 0.47 0 1 1162 It as a disadvantage if the municipality increased the extent

of the voluntary recycling programs (0,1) 0.22 0.42 0 1 1162